Icon Description4WARN Insights Blog

Who’s Influencing the AI Answers Your Policyholders Receive?

Insurance companies are spending a lot of time and money making sure the AI technology they use internally is responsible, accurate, and secure. But while they're focused on their own technology use, policyholders are accessing tools like ChatGPT, Claude, and Gemini every day to get answers about insurance coverage, claims, and what to do after an accident, crisis, or natural disaster.

But there's no way for insurers (or policyholders) to tell who or what is influencing those answers.

An insurer can do everything right when it comes to managing its own AI, but its policyholders may still get misleading information from websites and businesses that the insurer has no control over.

Who's Influencing AI Answers?

Most people know that search results can be manipulated, but what's changing is how that information reaches consumers. Someone looking for legitimate advice may now receive an AI generated answer without ever visiting the source websites behind it. If those sources contain misleading information, it can become part of an AI response that sounds credible. And there are businesses working hard to influence those answers.

A 2025 study by researchers at ETH Zurich demonstrated how easy it is to manipulate AI search tools. In one experiment, researchers created fictional camera products and tested whether carefully written website content could influence AI’s recommendations. BEFORE the manipulation started, AI recommended real, legitimate camera brands 57.9% of the time, and the fictional cameras only 34% of the time. AFTER researchers manipulated the website content, recommendations for the fictional products jumped to 59.4%, beating out the real-life brands.

What’s concerning about this research is that it demonstrates how easily online content can be deliberately manipulated to influence which products AI recommends, promote products that don’t even exist, or undermine legitimate competitors. When consumers trust those AI generated answers, they may never realize someone has influenced the information they’re relying on.

Now think about what happens when someone asks AI about an insurance claim. The answers in Gemini or ChatGPT may direct them to an attorney, a dishonest roofing contractor, or a business providing misleading medical advice. We already see businesses using websites, search results, and online ads to influence policyholders looking for help with their claims. Now that same misleading information can appear at the very top of search results in an AI generated response that looks reliable and trustworthy.

Protecting Policyholders Beyond Your Own AI Systems

Insurers can't control every website or AI generated answer, but there are practical ways to help policyholders find accurate information.

A Quick Checklist for Insurers

  1. Know what consumers are asking. What questions are people asking about your company, coverage, and claims?
  2. Check the answers. What information appears in search results and AI-generated responses? Are third parties appearing ahead of your company?
  3. Make accurate information easy for people and AI to find. Remember that you aren’t just publishing content for consumers to read. AI is now one of your primary target audiences. These tools are using your online content to answer policyholders’ questions every day. Be sure to publish useful FAQs, keep coverage and claims information current, make your phone number prominent on your website, and use legitimate SEO practices so reliable sources are easy for both consumers and AI tools to find.
  4. Monitor outside activity. Watch for suspicious websites, misleading information, and attempts to redirect policyholders to businesses that may profit from their decisions.

Insurers have spent years strengthening and protecting the technology, systems, and personal information they control. Now they need to pay just as much attention to what policyholders encounter outside those systems. Misleading information isn't just an inconvenience. It can send policyholders to fraudulent businesses, expose them to financial harm, interfere with legitimate claims, and undermine the trust they've placed in their insurer.

Monitoring this activity takes more than an occasional search or a review of your company’s website. 4WARN® scans thousands of websites to help insurers identify who’s targeting their policyholders, how they’re doing it, and what steps they can take to respond. To learn more about protecting your policyholders from online manipulation, contact the 4WARN team.

Source

Adversarial Search Engine Optimization for Large Language Models, ICLR 2025.

About the Author

John Dimond is a data-driven engineer focused on applying automation and advanced analytics to detect emerging risk patterns for a variety of industries. These efforts help shield organizations from “tech-enabled claim instigation” and promote sustainable innovation in the insurance, financial, transportation, healthcare, retail, and food services industries. In addition to his leadership role in data analytics, John also contributes across multiple executive functions at 4WARN, reflecting his broad expertise in corporate strategy and operations.

Before joining 4WARN, John developed his expertise in the insurance sector at DemoTech, Inc. where he offered strategic insights into the financial and regulatory aspects of risk management.

John holds a Bachelor of Applied Science in Electro-Mechanical Engineering degree from Miami University. Guided by a passion for technology and a commitment to innovation, he continues to shape the future of risk management through holistic, data-centric solutions.

Get in Touch with Us Today

Empower your strategy with 4WARN's expertise and steer clear of tech-enabled claim instigation risks.